Codes of Conduct

Ethical Data Science

Dr Zak Varty

Data Science: Miracle Cure and Sexiest Job


Data Science / ML / AI applications:

  • Health and Social care
  • Ecology and Conservation
  • Environmental Science
  • Business and Finance
  • Government and Policy Development

Title of Harvard Business Review article: Is data scientist still the sexiest job of the 21st Century?

What could go wrong?

BBC news article: facial recognition fails on race, government study says.

BBC news article: Uber's self-driving operator charged over fatal crash.

That’s not my type of data science …


Technological Adoption Relies on Public Trust


Other professions have strict codes of conduct and are legally liable for the outcomes of their work:


  • Doctors
  • Pharmaceutical companies
  • Lawyers
  • Civil and systems Engineers

Eight-lane roadbridge collapsed into the Mississippi river.

Thirteen people were killed and 145 injured during the Interstate 35W bridge collapse in 2007.

Hippocratic Oath

Bust of Hippocrates by an unidentified engraver. (Public domain via Wikimedia Commons).

Do No Harm:

Doctors around the world swear a version of the Hippocratic oath, originating in 400BCE, which introduced the principles of medical confidentiality and non-maleficence.

Corporate Oaths:

Google’s original internal code of conduct started with the phrase ‘Don’t be evil’. This was later rephrased to ‘You can make money without being evil’.

A Hippocratic Oath for Data Scientists


Wired article: To work for society, data scientists need a hippocratic oath with teeth

In Weapons of Math Destruction (2016), Cathy O’Neil was among the first authors to make a call for a Hippocratic oath for data scientists

Doing the right thing


Neither obvious or easy:

  • Lack of understanding,
  • Unanticipated consequences
  • Incentive structures,
  • Inherent trade-offs.

Sketch of a road sign pointing in two directions: right and wrong.

Codes of Conduct

Membership of a professional body can give accountability and support.

  • Codes of conduct
  • External accountability
  • External support (p2p and organisational)
  • Training and professional development

Currently no professional body dedicated to data science: RSS, OR Society, SIAM.

Wrapping Up

Data Science is still in it’s adolescence:

  • Widespread use means lots of potential for harm
  • Learning from the mistakes and successes of other disciplines.

Being associated with a professional body can help to:

  • Formalise data science as it’s own discipline
  • Incentivise good practices

We must remain alert to the potential dangers posed by our work.